CN108347611B - Optimization method of coding block-level Lagrange multiplier for theodolite - Google Patents

Optimization method of coding block-level Lagrange multiplier for theodolite Download PDF

Info

Publication number
CN108347611B
CN108347611B CN201810174851.6A CN201810174851A CN108347611B CN 108347611 B CN108347611 B CN 108347611B CN 201810174851 A CN201810174851 A CN 201810174851A CN 108347611 B CN108347611 B CN 108347611B
Authority
CN
China
Prior art keywords
longitude
theta
coding block
latitude
coding
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810174851.6A
Other languages
Chinese (zh)
Other versions
CN108347611A (en
Inventor
周益民
程学理
黄航
冷龙韬
王宏宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201810174851.6A priority Critical patent/CN108347611B/en
Publication of CN108347611A publication Critical patent/CN108347611A/en
Priority to PCT/CN2019/072822 priority patent/WO2019165863A1/en
Application granted granted Critical
Publication of CN108347611B publication Critical patent/CN108347611B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/154—Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/597—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding specially adapted for multi-view video sequence encoding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/189—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding
    • H04N19/19—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding using optimisation based on Lagrange multipliers
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/30—Image reproducers
    • H04N13/332—Displays for viewing with the aid of special glasses or head-mounted displays [HMD]
    • H04N13/344—Displays for viewing with the aid of special glasses or head-mounted displays [HMD] with head-mounted left-right displays

Abstract

The invention relates to a method for video codingThe method, especially for the technical field of encoding under VR360 video longitude and latitude map format, provides an optimization method of encoding block-level Lagrange multiplier for the longitude and latitude map, which calculates the ratio rho (theta) of the area of the spherical ring zone where the encoding block is located and the area of the longitude and latitude map pixel ring zone where the encoding block is located according to the position information of the encoding block in the longitude and latitude map, and according to the rho (theta) to lambdasysAnd optimizing calculation to obtain an optimized Lagrange multiplier lambda (rho (theta)), encoding the encoding block according to the lambda (rho (theta)), and introducing the position information of the encoding block into the correction and optimization of the block-level Lagrange multiplier in an area ratio mode, so that the overall performance of the longitude and latitude map encoding is remarkably improved, and the method is suitable for video encoding under a VR360 video longitude and latitude map format.

Description

Optimization method of coding block-level Lagrange multiplier for theodolite
Technical Field
The invention relates to a video coding method, in particular to an optimization method of a coding block-level Lagrangian multiplier for a longitude and latitude map, which is also called an EquiRectangular Projection (ERP) map for short, aiming at the technical field of coding under a VR360 video longitude and latitude map format.
Background
Virtual Reality (VR) technology is a computer simulation system that creates and experiences an immersive Virtual world. The method integrates the latest development of technologies such as computer graphics, computer simulation, artificial intelligence, induction, display, network parallel processing and the like. VR technology is typically generated with the aid of computer technology, and is often in the form of a simulated virtual display system. With the rapid development of VR technology, consumer electronics related to VR gradually come into people's lives. Currently, most VR content is oriented to the visual experience. Typically presented via a computer screen, a special display device or a stereoscopic display device. VR technology application scenes are reflected in the game industry and the movie industry, and a large number of VR game products and VR video contents are continuously put into the market in recent years. More widely, VR has a large number of applications in the fields of medicine, education, aerospace, rail transit, and the like. VR technology has become a current area of research.
In order to enhance the experience of the user, the source parameters such as the resolution, the pixel representation range, the frame rate, etc. of the VR video image are generally significantly higher than those of the normal video, mainly 8K and 4K. Compared with high-definition 1080P video, the data volume is improved by dozens of times. Therefore, how to continuously improve the compression efficiency of VR video by technical means becomes a new technical challenge.
Rate Distortion Optimization (RDO) is the most critical core optimization technique in video coding, and is supported by Rate Distortion theory. The rate distortion optimization technology can solve the generation problem of the optimized code stream of the encoder, and is ensured by a rate distortion optimization theory. The basic problems of rate-distortion theory are: the rate-distortion optimization technique for video coding achieves the minimum desired distortion for a given source distribution and distortion metric at a particular code rate.
In a particular application, rate-distortion optimization translates the problem into selecting a set of parameters in a given set of encoding parameters, so that video can be encoded at the least bit rate under defined distortion conditions. The theoretical optimal coding parameters can be obtained by traversing all the optional coding parameter sets by adopting an exhaustive method, but the time complexity of the operation of the exhaustive method is extremely high, the time required by coding is extremely long, and the method can not be applied to actual coding. Meanwhile, as the video coding is performed by taking the coding units as units and the parameters of each coding unit are independent of each other, the optimal coding parameters of each coding unit can be considered to belong to the optimal coding parameter set of the whole coding process, i.e. the global optimal problem is decomposed into a set of a plurality of local optimal problems.
The rate distortion optimization process introduces a lagrange multiplier lambda (lambda) to transform the unconstrained optimization problem into a constrained optimization problem. Since the Lagrange optimization method is introduced to solve the rate distortion optimization problem, the video coding rate distortion optimization technology has practical application value. Due to the lower complexity and higher performance, the method is widely popularized. Currently, rate-distortion optimization techniques based on lagrangian multipliers are now being applied to mainstream h.264/AVC and HEVC/h.265 encoders. In general, the value of λ is determined by means of a high bit hypothesis derivation formula, and in practical use, an empirical value is added for correction according to different encoder characteristics. The quality of the selected lambda value is directly related to the coding performance.
The quality of video coding is evaluated and generally described by using BD-RATE and BD-PSNR, and the description method can be found in the following documents: [ Gisle Bjontegaard, conservation of Average PSNR Differences between RD curves, ITU-T SC16/Q6,13th VCEG Meeting, Austin, Texas, USA, April 2001, doc. VCEG-M33 ]. The calculation processes of the two are similar, and the integral difference operation is carried out on the basis of carrying out high-order interpolation connection by acquiring the objective quality PSNR and the coding Bit Rate (Bit-Rate) of the test point. In general, the bit rate is statistically normalized and unambiguous, but common two-dimensional video and VR360 theodolite video are not much the same in terms of objective quality PSNR.
Common two-dimensional video coding generally uses Peak Signal to Noise Ratio (PSNR) as an objective quality evaluation index. And because the VR360 video sequence is usually stored in a storage medium in the form of a longitude and latitude map, the VR360 video sequence is projected and mapped into a spherical surface during playing, thereby presenting the effect of 360-degree stereo surrounding. The pixel compression effect is inevitably generated in the process of mapping from the longitude and latitude map to the spherical surface. That is, pixels at the same latitude except the equator are compressed when they are mapped to the spherical surface, and the higher the latitude, the more intense the compression. Under extreme conditions, a row of pixel points of south and north poles in the longitude and latitude map are compressed into a pixel point of two poles of a sphere. Therefore, in view of the particularity of the 360VR video longitude and latitude representation format, the 360VR video longitude and latitude representation format is not directly displayed in the playing process, but is synthesized on the spherical surface and then output and displayed, so that the objective quality of the three-dimensional spherical surface cannot be accurately described by using the two-dimensional PSNR.
Therefore, the experts in the field propose improved objective evaluation models such as spherical Peak Signal to Noise Ratio (SPSNR), Weighted spherical Peak Signal to Noise Ratio (wspspsnr), krarsted Parabolic Projection Peak Signal to Noise Ratio (CPP-PSNR), etc. as the 360VR video objective evaluation indexes which are relatively common at present, wherein the SPSNR is subdivided into a spherical Peak Signal to Noise Ratio with Interpolation (SPSNR-I) and a Nearest Neighbor spherical Peak Signal to Noise Ratio (SPSNR-NN).
Therefore, it is worth noting that the existing video encoders are designed for general two-dimensional images, and the source property of VR360 longitude and latitude map format is not considered specially, which may cause the performance loss of SPSNR or WSPSNR to be serious even though the PSNR performance is kept good.
Disclosure of Invention
The invention provides an optimization method of a coding block-level Lagrange multiplier for a longitude and latitude map, which can optimize the Lagrange multiplier and is beneficial to improving the overall performance of the longitude and latitude map coding.
The invention discloses an optimization method of a coding block-level Lagrange multiplier for a longitude and latitude map, which comprises the following steps:
A. acquiring 1 frame image of a video sequence;
B. sequentially obtaining 1 coding block in a current frame;
C. b, calculating the ratio rho (theta) of the area of the annular zone of the spherical surface where the coding block is positioned to the area of the pixel annular zone of the longitude and latitude image where the coding block is positioned according to the position information of the coding block in the longitude and latitude image obtained in the step B, wherein the theta is a calculated value of the zenith angle of the coding block in the spherical surface;
D. according to rho (theta) to lambdasysPerforming optimization calculation to obtain optimized Lagrange multiplier lambda (rho (theta)), and the lambda issysB, obtaining a Lagrange multiplier subsystem value of the current frame in the step A;
E. coding the coding block according to the lambda (rho (theta)) obtained in the step D;
F. judging whether all the coding blocks in the current frame are coded, if so, entering a step G, otherwise, entering a step B;
G. and C, judging whether the full sequence coding is finished after the current frame coding is finished, if so, finishing, and otherwise, continuing to code in the step A.
Further, λ (ρ (θ)) ═ λ in step Dsys·(ξ+ρ(θ))γWherein λ issysIs a Greenian multiplier subsystem value, theta is a calculated value of the zenith angle of the current coding block in the spherical surface, xi is a minimum value for preventing the zero division operation,
Figure BDA0001586986520000031
is a model parameter related to the image content and β is a model parameter related to the source characteristics.
Specifically, in step C, the calculated zenith angle of the coding block in the spherical surface is θ, wherein,
area S of spherical annulusspher(θ) is represented by the formula: sspher(θ)＝2π·r·sinθ·hringIs obtained by calculation, wherein hringIs the height of the spherical annulus, where r is the radius of the sphere;
area S of pixel ring of longitude and latitude imageerp(θ) is represented by the formula:
Figure BDA0001586986520000032
and (6) calculating.
Further, the height h of the ring beltringR · sin d θ, the coding block,
area S of spherical annulusspher(θ) is represented by the formula: sspher(θ)＝2π·r2Sin theta sin d theta is obtained by calculation,
area S of pixel ring of longitude and latitude imageerp(θ) is represented by the formula:
Figure BDA0001586986520000033
the calculation results are that,
area ratio in step C
Figure BDA0001586986520000034
The above d θ is an angle difference of a zenith angle formed by the upper edge and the lower edge of the endless belt.
Preferably, the step of obtaining the zenith angle calculated value θ in the step C includes:
C1. the coordinate position on the longitude and latitude map where the current coding block is located is expressed as: the subscript of the first row of the current coding block in the whole longitude and latitude map is k, the pixel height of the coding block is N, and the total pixel height of the longitude and latitude map is h;
C2. according to the data obtained in the step C1, the zenith angle corresponding to the pixel with the row index i in the current coding block is θ (i), and the vertex angle is θ (i)
Figure BDA0001586986520000041
By the formula
Figure BDA0001586986520000042
Calculating to obtain the arithmetic average value of the zenith angles theta (i) of the pixels of each row of the current coding block
Figure BDA0001586986520000043
And averaging the arithmetic mean
Figure BDA0001586986520000044
As a zenith angle calculation value θ in ρ (θ), obtained
Figure BDA0001586986520000045
Specifically, in step a, the position of the image in the sequence is determined, and the frame type, the frame attribute, and the position and the level in the picture group in which the image is located are determined; and according to the obtained frame attribute of the current frame, calculating by an encoder to obtain a Lagrange multiplier subsystem value lambda of the frame levelsys。
The invention has the beneficial effects that:
currently, the objective quality evaluation for 360VR video is still based on the second order distance (MSE) of the conventional distorted pixel Error. The distortion calculation process on the VR360 longitude and latitude map is not like point-to-point MSE statistics of a 2D image, but is put on a 3D spherical surface to perform average calculation in the meaning of effectively representing area equivalence. Obviously, rate distortion optimization on VR360 theodolite should be modified accordingly to fit the distortion calculation rule of the sphere. Because the VR360 longitude and latitude map distortion calculation is the distortion accumulation of the same area on the sphere, it is necessary to explain the ratio of pixels CTUs (coded blocks) at different latitudes in the final SPSNR (sphere peak signal-to-noise ratio) calculation according to the analysis of the longitude and latitude map to sphere mapping process.
As is known in the art, the longitude mapping from the sphere to the VR360 longitude map is scaled, and the latitude mapping is a direct projection process from the sphere to the cylinder. Then the ratio of the area of the spherical annulus to the area of the pixels of the VR360 longitude and latitude lines relates only to the latitude and not to the longitude.
While lagrange multipliers are usually expressed as functions that are closely related to the quantization step size. Various encoding platforms have different parameter correction factors for lagrange multipliers to be close to the R-D curves thereof to achieve the highest possible encoding gain.
The invention creatively constructs the weight by the ratio of the area of the spherical ring zone where the coding block is positioned to the area of the longitude and latitude image pixel ring zone where the coding block is positioned, introduces the position information of the coding block in the form of the area ratio, then uses the weight containing the position information to correct and optimize the coding block-level Lagrange multiplier, and finally uses a new quantization parameter to code, thereby obviously improving the overall performance of the longitude and latitude image coding.
Drawings
Fig. 1 is a schematic diagram of a mapping relationship between a VR360 video longitude and latitude map and spherical pixels.
Fig. 2 is a schematic view of a spherical projection of a VR360 longitude and latitude map.
Fig. 3 is a front view of fig. 2.
Fig. 4 is a right side view of fig. 2.
Fig. 5 is a rear view of fig. 2.
Fig. 6 is a left side view of fig. 2.
Fig. 7 is a top view of fig. 2.
Fig. 8 is a bottom view of fig. 2.
Fig. 9 is a flowchart of the optimization method for the encoding block-level lagrangian multiplier for the longitude and latitude map according to the present invention.
Fig. 10 is a flow chart of the optimization of the values of the lagrangian multiplier subsystem of fig. 9.
Detailed Description
As shown in fig. 1 to 8, the method for optimizing the encoding block-level lagrangian multiplier for the longitude and latitude map according to the present invention includes the following steps:
A. acquiring 1 frame image of a video sequence;
B. sequentially obtaining 1 coding block in a current frame;
C. b, calculating the ratio rho (theta) of the area of the annular zone of the spherical surface where the coding block is positioned to the area of the pixel annular zone of the longitude and latitude image where the coding block is positioned according to the position information of the coding block in the longitude and latitude image obtained in the step B, wherein the theta is a calculated value of the zenith angle of the coding block in the spherical surface;
D. according to rho (theta) to lambdasysPerforming optimization calculation to obtain optimized Lagrange multiplier lambda (rho (theta)), and the lambda issysB, obtaining a Lagrange multiplier subsystem value of the current frame in the step A;
E. coding the coding block according to the lambda (rho (theta)) obtained in the step D;
F. judging whether all the coding blocks in the current frame are coded, if so, entering a step G, otherwise, entering a step B;
G. and C, judging whether the full sequence coding is finished after the current frame coding is finished, if so, finishing, and otherwise, continuing to code in the step A.
Generally, the lagrange multiplier is generally trained by a large amount of experimental data, is calculated by an empirical formula, and is expressed as a function closely related to the quantization step size. Various encoding platforms have different parameter correction factors for the Lagrangian multiplier to be close to the R-D curve of the frames so as to obtain the highest possible encoding gain, and therefore the Lagrangian multiplier system definition value lambda of the frames can be obtainedsys。
Furthermore, the invention constructs the weight by the ratio lambda (rho (theta)) of the area of the spherical ring zone where the coding block is located and the area of the pixel ring zone of the longitude and latitude map where the coding block is located, introduces the position information of the coding block in the form of the area ratio, then uses the weight containing the position information to correct and optimize the coding block-level Lagrange multiplier, and finally uses a new quantization parameter to code, thereby obviously improving the overall performance of the longitude and latitude map coding.
The formula for calculating λ (ρ (θ)) in step D may be constructed differently depending on the area ratio weight structure, the optimization purpose, and the like. In this embodiment, specifically, in step D
λ(ρ(θ))＝λsys·(ξ+ρ(θ))γ (1)
Wherein λ issysIs a Greenian multiplier subsystem value, theta is a calculated value of the zenith angle of the current coding block in the spherical surface, xi is a minimum value for preventing the zero division operation,
Figure BDA0001586986520000061
is a model parameter related to the image content and β is a model parameter related to the source characteristics.
The derivation process of the above equation (1) is as follows:
and (4) calculating the SPSNR (spherical peak signal-to-noise ratio), and sampling according to the mapping pixel density. Specifically, the number of sampling points is large at a position with a low latitude, that is, a high pixel density; the number of sampling points is small at a position with high latitude, namely low pixel density. Based on the principle that the bit number allocated to the VR360 longitude and latitude map is consistent with the area proportion of the corresponding spherical zone, the bit number consumed by coding is reduced while the subjective quality and the objective quality are maintained.
Considering that the area ratio is a sine function of a zenith angle calculated value theta, and the distribution of the longitude and latitude map bit rate is expected to meet the requirement of spherical display, a ratio model is established through a formula (2):
Figure BDA0001586986520000062
wherein the content of the first and second substances,
Figure BDA0001586986520000063
and R (θ) is the coding bit rate calculated as the θ zone at the equator and zenith angles, respectively. Based on the model given in equation (2), the spherical surface λ can be derived from the following steps:
the R-lambda model of the relationship between bit rate and Lagrange multiplier is
R＝α·λβ (3)
Where α and β are model parameters related to source characteristics.
Substituting equation (3) into equation (2) yields equation (4)
Figure BDA0001586986520000071
Wherein, λ (θ) and
Figure BDA0001586986520000072
respectively, the values of a sphere lambda with the calculated value of theta at the equator and the zenith angle. From equation (4), the proportional equation of the spherical surface λ can be obtained by sorting, as shown in equation (5):
Figure BDA0001586986520000073
wherein the content of the first and second substances,
Figure BDA0001586986520000074
are model parameters that are related to the image content.
Thus, according to the derivation process described above, equation (1) is established.
According to the geometric calculation, in step C, the vertex angle calculation value of the coding block in the spherical surface is theta, wherein,
area S of spherical annulusspher(θ) is represented by the formula: sspher(θ)＝2π·r·sinθ·hringIs obtained by calculation, wherein hringIs the height of the spherical annulus, where r is the radius of the sphere;
area S of pixel ring of longitude and latitude imageerp(θ) is represented by the formula:
Figure BDA0001586986520000075
and (6) calculating.
For further convenience of calculation, the height h of the annulusringR · sin d θ, the coding block,
area S of spherical annulusspher(θ) is represented by the formula: sspher(θ)＝2π·r2Sin theta sin d theta is obtained by calculation,
area S of pixel ring of longitude and latitude imageerp(θ) is represented by the formula:
Figure BDA0001586986520000076
the calculation results are that,
area ratio in step C
Figure BDA0001586986520000081
The above d θ is an angle difference of a zenith angle formed by the upper edge and the lower edge of the endless belt.
Of course, according to the difference between the area ratio calculation process and the value taking mode, ρ (θ) is not necessarily equal to sin θ, but does not affect the implementation of the present invention, and only affects the calculation difficulty of the implementation process of the present invention.
The above θ is called zenith angle calculation, not zenith angle, because:
as shown in fig. 1, VR360 is a mapping relationship between a graticule Projection (ERP) and spherical pixels, i.e. a mapping relationship between a cylindrical Projection (ERP) and a graticule map, and the mapping relationship is an angle in longitude
Figure BDA0001586986520000088
And expressing the latitude by using a zenith angle theta, wherein the latitude included angle corresponding to the upper boundary and the lower boundary of the spherical zone pixel is d theta. The d theta is a determined value on the premise of determining the spherical ring bandwidth, is influenced by the spherical ring bandwidth, and is not unique in each row of pixels contained in the annular band, so that theta is called a zenith angle calculated value for convenient calculation, the value of theta can be an extreme value, a specific value, an average value of the maximum and minimum values, an arithmetic average value and the like, and the value can be specifically taken according to the division rule of the coding block, the optimization requirement and the like.
In this embodiment, the zenith angle calculated value θ is taken as an arithmetic average value, and therefore, the step of obtaining the zenith angle calculated value θ in the step C includes:
C1. the coordinate position on the longitude and latitude map where the current coding block is located is expressed as: the subscript of the first row of the current coding block in the whole longitude and latitude map is k, the pixel height of the coding block is N, and the total pixel height of the longitude and latitude map is h;
C2. according to the data obtained in the step C1, the zenith angle corresponding to the pixel with the row index i in the current coding block is θ (i), and the vertex angle is θ (i)
Figure BDA0001586986520000082
By the formula
Figure BDA0001586986520000083
Calculating to obtain the arithmetic average value of the zenith angles theta (i) of the pixels of each row of the current coding block
Figure BDA0001586986520000084
And averaging the arithmetic mean
Figure BDA0001586986520000085
As a zenith angle calculation value θ in ρ (θ), obtained
Figure BDA0001586986520000086
Therefore, the formula (7) is substituted into the formula (1), and the final optimized calculation formula of the embodiment is obtained as
Figure BDA0001586986520000087
As shown in fig. 9, a specific block diagram of the optimization method of the encoding block-level lagrangian multiplier for a longitude and latitude map in this embodiment is shown, and fig. 10 is a specific block diagram of the calculation of λ (ρ (θ)) in this embodiment, and the whole optimization process steps are as follows:
1) acquiring 1 frame image of a video sequence;
2) determining the position of the obtained current frame in the sequence, and determining the frame type, the frame attribute, and the position and the layer in the picture group;
3) according to the obtained frame attribute of the current frame, calculating by an encoder to obtain a Lagrange multiplier subsystem value lambda of the frame levelsys；
4) Sequentially obtaining 1 coding block in a current frame, and determining the position information of the coding block in a longitude and latitude chart, wherein the position information is expressed as: the subscript of the first row of the current coding block in the whole longitude and latitude map is k, the pixel height of the coding block is N, and the total pixel height of the longitude and latitude map is h;
5) according to the position information of the coding block in the longitude and latitude map, the zenith angle corresponding to the pixel with the line subscript i in the current coding block is theta (i), and the zenith angle is theta (i)
Figure BDA0001586986520000091
By the formula
Figure BDA0001586986520000092
Calculating to obtain the arithmetic average value of the zenith angles theta (i) of the pixels of each row of the current coding block
Figure BDA0001586986520000093
6) According to rho (theta) to lambdasysPerforming optimization calculation to obtain an optimized Lagrange multiplier lambda (rho (theta)), wherein the formula is as follows:
Figure BDA0001586986520000094
7) encoding the encoded block according to the calculated λ (ρ (θ));
8) judging whether all the coding blocks in the current frame are coded, if so, entering a step 9), and otherwise, entering a step 4);
9) and judging whether the coding of the full sequence is finished after the coding of the current frame is finished, if so, finishing, and otherwise, turning to the step 1) to continue coding.
Table 1 shows the test sequence of the joint virtual reality standard group (IEEE 1857.9VRU) of the international institute of electrical and electronics engineers 1857.9 topic group, and table 2 and table 3 show the performance gain of the present invention compared to the existing unoptimized lagrangian multiplier test under two different test configuration conditions, respectively. ξ above is a minimum value that prevents the divide-by-zero operation, so it is as small as possible; gamma and beta are model parameters related to image content and related to information source characteristics, and in the embodiment, xi takes an empirical value of 0.015 and takes a value of 0.20 according to test data gamma.
Sequence Name Resolution Frame Count FPS Bit depth
Fengjing1 4096×2048 300 30 8
Fengjing3 4096×2048 300 30 8
Hangpai1 4096×2048 300 30 8
Hangpai2 4096×2048 300 30 8
Hangpai3 4096×2048 300 30 8
Xinwen1 4096×2048 300 30 8
Xinwen2 4096×2048 300 30 8
TABLE 1 theodolite video test sequence
The test set is a universal 7-piece 4K VR360 longitude and latitude map video test sequence, and the test adopts full sequence test. The performance index is performed by the BD-RATE performance statistics common in the art, with negative values indicating the saved bit RATE ratio at equivalent objective quality and positive values indicating the wasted bit RATE ratio at equivalent objective quality. Negative values of the BD-RATE generally indicate the degree of gain of the algorithm. The test baseline was based on AVS2 (chinese second generation audio video standard) universal test conditions. The Test results of the four QP points 27, 32, 38, and 45 in the default configuration of the system are used as comparison basis (Anchor), the four Test results at the same code RATE point are used as Test basis (Test) in the method of this embodiment, and BD-RATE is respectively calculated and counted under two conditions of traditional PSNR (peak signal to noise ratio) and longitude and latitude map SPSNR (spherical peak signal to noise ratio). The test is divided into two typical configuration structures of Low Delay (LD) and Random Access (RA).
Figure BDA0001586986520000101
Table 2, experimental results of the optimization method of the present invention in a Low Delay (LD) configuration.
In table 2, the gain degree of objective quality under two different evaluation modes of traditional PSNR and SPSNR under the configuration of Low Delay (LD) is counted. We can clearly see that the gain in SPSNR reaches 2.8%, higher than the 0.4% gain in PSNR-Y. Particularly, on the sequence Fengjing1, the gain of PSNR-Y reaches 3.3 percent, and the gain of SPSNR reaches 5.8 percent; the gain of SPSNR is even more 6.4% over the Hangpai1 sequence.
Figure BDA0001586986520000111
TABLE 3 Experimental results of the optimization method of the present invention in Random Access (RA) configuration
In table 3, the gain degree of the objective quality under two different evaluation modes, i.e., PSNR and SPSNR, in the Random Access (RA) configuration is counted. We can clearly see that the gain in SPSNR reaches 1.5%. In particular, gains of three sequences, Hangpai1, Hangpai2 and Hangpai3, were as high as 2.8%, 2.1% and 2.3%, respectively, and an average gain of 2.4% was obtained.
The test results in tables 2 and 3 clearly show that the optimization method of the present invention can significantly optimize the coding block level lagrangian multiplier for VR360 video image coding, and significantly improve the video coding efficiency in both configurations.

Claims (5)

1. The optimization method of the encoding block-level Lagrange multiplier for the longitude and latitude map comprises the following steps:
A. acquiring 1 frame image of a video sequence;
B. sequentially obtaining 1 coding block in a current frame;
C. b, calculating the ratio rho (theta) of the area of the annular zone of the spherical surface where the coding block is positioned to the area of the pixel annular zone of the longitude and latitude image where the coding block is positioned according to the position information of the coding block in the longitude and latitude image obtained in the step B, wherein the theta is a calculated value of the zenith angle of the coding block in the spherical surface;
D. according to rho (theta) to lambdasysPerforming optimization calculation to obtain an optimized Lagrange multiplier lambda (rho (theta)):
λ(ρ(θ))＝λsys·(ξ+ρ(θ))γ；
wherein λ issysThe Lagrange multiplier subsystem value of the current frame obtained in the step A is obtained, theta is the calculated value of the zenith angle of the current coding block in the spherical surface, xi is the minimum value for preventing the zero division operation,
Figure FDA0002747038490000011
is a model parameter related to the image content, β is a model parameter related to the source characteristics;
E. coding the coding block according to the lambda (rho (theta)) obtained in the step D;
F. judging whether all the coding blocks in the current frame are coded, if so, entering a step G, otherwise, entering a step B;
G. and C, judging whether the full sequence coding is finished after the current frame coding is finished, if so, finishing, and otherwise, continuing to code in the step A.
2. The method for optimizing an encoded block-level lagrange multiplier for theodolites as claimed in claim 1, characterized by: in step C, the vertex angle calculation value of the coding block in the spherical surface is theta, wherein,
area S of spherical annulusspher(θ) is represented by the formula: sspher(θ)＝2π·r·sinθ·hringIs obtained by calculation, wherein hringIs the height of the spherical annulus, where r is the radius of the sphere;
area S of pixel ring of longitude and latitude imageerp(θ) is represented by the formula:
Figure FDA0002747038490000012
and (6) calculating.
3. The method for optimizing an encoded block-level lagrange multiplier for theodolites as claimed in claim 2, characterized by:
height h of the endless beltringWhen r.sind theta, the coding block,
area S of spherical annulusspher(θ) is represented by the formula: sspher(θ)＝2π·r2Sin theta and sind theta are obtained by calculation,
area S of pixel ring of longitude and latitude imageerp(θ) is represented by the formula:
Figure FDA0002747038490000013
the calculation results are that,
area ratio in step C
Figure FDA0002747038490000021
The above d θ is an angle difference of a zenith angle formed by the upper edge and the lower edge of the endless belt.
4. The method for optimizing an encoded block-level lagrange multiplier for theodolites as claimed in claim 3, characterized by: the step of obtaining the zenith angle calculated value θ in the step C includes:
C1. the coordinate position on the longitude and latitude map where the current coding block is located is expressed as: the subscript of the first row of the current coding block in the whole longitude and latitude map is k, the pixel height of the coding block is N, and the total pixel height of the longitude and latitude map is h;
C2. according to the data obtained in the step C1, the zenith angle corresponding to the pixel with the row index i in the current coding block is θ (i), and the vertex angle is θ (i)
Figure FDA0002747038490000022
By the formula
Figure FDA0002747038490000023
Calculating to obtain the arithmetic average value of the zenith angles theta (i) of the pixels of each row of the current coding block
Figure FDA0002747038490000024
And averaging the arithmetic mean
Figure FDA0002747038490000025
As a zenith angle calculation value θ in ρ (θ), obtained
Figure FDA0002747038490000026
5. The method for optimizing an encoded block-level lagrange multiplier for theodolites as claimed in claim 3, characterized by:
in step A, determining the position of the image in the sequence, and determining the frame type, the frame attribute, the position and the layer of the image group; and according to the obtained frame attribute of the current frame, calculating by an encoder to obtain a Lagrange multiplier subsystem value lambda of the frame levelsys。
CN201810174851.6A 2018-03-02 2018-03-02 Optimization method of coding block-level Lagrange multiplier for theodolite Active CN108347611B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201810174851.6A CN108347611B (en) 2018-03-02 2018-03-02 Optimization method of coding block-level Lagrange multiplier for theodolite
PCT/CN2019/072822 WO2019165863A1 (en) 2018-03-02 2019-01-23 Optimization method for encoding-block-level lagrangian multiplier of equirectangular image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810174851.6A CN108347611B (en) 2018-03-02 2018-03-02 Optimization method of coding block-level Lagrange multiplier for theodolite

Publications (2)

Publication Number Publication Date
CN108347611A CN108347611A (en) 2018-07-31
CN108347611B true CN108347611B (en) 2021-02-02

Family

ID=62958947

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810174851.6A Active CN108347611B (en) 2018-03-02 2018-03-02 Optimization method of coding block-level Lagrange multiplier for theodolite

Country Status (2)

Country Link
CN (1) CN108347611B (en)
WO (1) WO2019165863A1 (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108347611B (en) * 2018-03-02 2021-02-02 电子科技大学 Optimization method of coding block-level Lagrange multiplier for theodolite
WO2020172813A1 (en) * 2019-02-27 2020-09-03 Oppo广东移动通信有限公司 Rate distortion optimization method and apparatus, and computer-readable storage medium
CN109889837B (en) * 2019-03-12 2022-05-27 电子科技大学 Reference fixed point calibration method for optimal Lagrange multiplier
CN115022638A (en) * 2022-06-30 2022-09-06 电子科技大学 Rate distortion optimization method for panoramic video coding

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103873867A (en) * 2014-03-31 2014-06-18 清华大学深圳研究生院 Free viewpoint video depth map distortion prediction method and free viewpoint video depth map coding method
WO2017051072A1 (en) * 2015-09-23 2017-03-30 Nokia Technologies Oy A method, an apparatus and a computer program product for coding a 360-degree panoramic video
CN107135397A (en) * 2017-04-28 2017-09-05 中国科学技术大学 A kind of panorama video code method and apparatus

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012050808A1 (en) * 2010-09-29 2012-04-19 Dolby Laboratories Licensing Corporation Region based asymmetric coding for 3d video compression
CN108347611B (en) * 2018-03-02 2021-02-02 电子科技大学 Optimization method of coding block-level Lagrange multiplier for theodolite

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103873867A (en) * 2014-03-31 2014-06-18 清华大学深圳研究生院 Free viewpoint video depth map distortion prediction method and free viewpoint video depth map coding method
WO2017051072A1 (en) * 2015-09-23 2017-03-30 Nokia Technologies Oy A method, an apparatus and a computer program product for coding a 360-degree panoramic video
CN107135397A (en) * 2017-04-28 2017-09-05 中国科学技术大学 A kind of panorama video code method and apparatus

Also Published As

Publication number Publication date
WO2019165863A1 (en) 2019-09-06
CN108347611A (en) 2018-07-31

Similar Documents

Publication Publication Date Title
US11783512B2 (en) Attribute value of reconstructed position associated with plural original points
CN108347611B (en) Optimization method of coding block-level Lagrange multiplier for theodolite
KR102594003B1 (en) Method, apparatus and stream for encoding/decoding volumetric video
KR102537087B1 (en) Motion estimation using 3D auxiliary data
Merkle et al. The effects of multiview depth video compression on multiview rendering
Lei et al. Depth coding based on depth-texture motion and structure similarities
US11838547B2 (en) Method and device for encoding the geometry of a point cloud
CN111386556A (en) Method and apparatus for generating points of a 3D scene
US20210211703A1 (en) Geometry information signaling for occluded points in an occupancy map video
EP3721414A1 (en) A method and apparatus for encoding/decoding the geometry of a point cloud representing a 3d object
TW202147842A (en) Point cloud quality assessment method, encoder, decoder and storage medium not only simplifying the computational complexity of subjective quality assessment, but also improving the accuracy of subjective quality assessment
CN106303521B (en) A kind of HEVC Rate-distortion optimization method based on sensitivity of awareness
Lei et al. Region adaptive R-$\lambda $ model-based rate control for depth maps coding
KR20180131454A (en) Method and apparatus for processing a video signal
CN104244008B (en) Depth video coding method
AU2020101803A4 (en) Method for Optimizing the Coding Block-Level Lagrange Multiplier of Longitude-Latitude Image
CN117280680A (en) Parallel mode of dynamic grid alignment
CN105007494A (en) Intra-frame wedge-shaped segmentation mode selection method of 3D video depth image model mode
JP7376211B2 (en) Signaling camera parameters in point cloud coding
CN103997653A (en) Depth video encoding method based on edges and oriented toward virtual visual rendering
RU2782251C1 (en) Methods and device for generalized encoding of geometry with a triangular grid
US20230412849A1 (en) Mesh vertex displacements coding
US20230177736A1 (en) Method and apparatus for chart based mesh compression
Javaheri Point Cloud Quality Assessment
Maceira et al. Region-based depth map coding using a 3D scene representation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant